1. Field of the Invention
The present invention relates to a technique for encoding and decoding an image signal, and more particularly, to an image encoding device, an image decoding device, an encoding method, and an image decoding method using a prediction processing in a screen which generates a prediction signal for an image signal as a target from an encoded image and encodes a differential signal with the prediction signal.
2. Description of the Related Art
In recent years, a service of delivering digitalized image and voice contents via a broadcast wave of a satellite or ground wave or a network has been put to practical use and high-efficiency encoding for efficiently recording information of an image and voice having a large amount of information has bee required. With high definition and diversification of contents, the number of images and the amount of data which are encoded have been further increased and high-efficiency encoding of an advanced image signal of a future content service has been required.
As high-efficiency encoding of an image, a method of compressing the amount of information by using a correlation between pixels which are adjacent spatially in the same frame of a moving image signal and a correlation between frames or fields which are adjacent temporally, which is representative as MPEG2 or MPEG4-AVC (Advanced Video Coding), has been used.
In the MPEG4-AVC (Literature ‘ISO/IEC14496-10 Advanced Video Coding’), the image is divided into a plurality of 2D blocks, the prediction signal is generated by using the correlation between the pixels in the same frame or between frames per block, and differential information is encoded together with the prediction signal, thereby implementing high encoding efficiency. Prediction processing using the correlation between the pixels in the same frame in the MPEG4-AVC is called intra-prediction and as illustrated in FIG. 20, a prediction image of a block as an encoding target is generated by using a decoded image of an encoded part adjacent to the target block. In intra-prediction, an adjacent decoding image is encoded together with the prediction mode information by selecting a prediction mode that has a smallest error from the encoding target block from the prediction image generated in a plurality of (9 types when prediction is performed per 4×4 pixel block) prediction modes having a high correlation with respect to a predetermined direction as illustrated in FIG. 20.
The intra-prediction becomes prediction processing using only the correlation with an adjacent area and in the case where the correlation is small on a boundary with the encoding target block, a prediction effect is reduced.
In the image signal, a signal having a similar image component such as a form, a shape, or a background of an object is present, even at a position separated from the target block. Japanese Patent Application Laid-open No. 2005-159947 discloses a method of performing prediction processing by using a correlation with an image at a position separated from a target block. In detail, as illustrated in FIG. 21, an error between an encoded decoding image and the encoding target block is calculated at a position which is shifted by a disparity amount (hereinafter, referred to as a disparity vector) in a screen from the target block, and a reference image referred as the disparity vector having a smallest error is referred to as the prediction image and encoded together with the disparity vector.
In the case where a prediction image having a high correlation over a decode amount of transmitting the disparity vector may be generated, high encoding efficiency can be implemented with respect to the intra-prediction, but in the case where a transmission quantity of the disparity vector is large, sufficient prediction efficiency cannot be achieved.
Japanese Patent Application Laid-open No. 2007-043651 discloses a method of specifying the disparity vector without transmitting the disparity vector in order to reduce the decode amount required for the disparity vector. In Japanese Patent Application Laid-open No. 2007-043651, by using an encoded decoding image adjacent to the target block as a template, an error between the encoded decoding image at a position which is shifted by the disparity vector and the encoded decoding image adjacent to the target block is calculated, a disparity vector having a smallest error is judged as a disparity vector of the encoding target block, and a reference image referred as the disparity vector is referred to as the prediction image. In this method, like the encoding side, in the decoding side, since the disparity vector can be calculated without being received by detecting the disparity vector with the encoded decoding image, the encode amount is not increased due to additional information.    [Patent Document 1] Japanese Patent Application Laid-open No. 2005-159947    [Patent Document 2] Japanese Patent Application Laid-open No. 2007-043651
In MPEG4-AVC, high encoding efficiency can be achieved by performing motion compensation prediction processing based on decoded image signals of different frames in a temporal direction with respect to image signals which are temporally consecutive, but a reference frame to perform motion compensation prediction processing needs to be encoded by using only intra-prediction in the same frame and when a correlation is low on a boundary with an encoding target block, a prediction effect is reduced. Even when a moving-image is not temporally consecutive, the motion compensation prediction is not functioned, and as a result, encoding efficiency is deteriorated by a limit in performance of the intra-prediction.
As presented in Japanese Patent Application Laid-open Nos. 2005-159947 and 2007-043651, a disparity vector from a target block is transmitted or self-generated, such that since prediction using self-similarity of a part other than an adjacent area to an image signal or prediction from different objects having the same texture component is performed when an image correlation at a position separated from the target block is used, self-similarity of an image signal or similarity of a texture cannot be significantly used due to shape disparity or luminance disparity of an object which is present in a screen.
In Japanese Patent Application Laid-open No. 2005-159947, when a prediction signal having an effect to exceed a code amount of the disparity vector cannot be calculated, efficiency deteriorates and in Japanese Patent Application Laid-open No. 2007-043651, since the disparity vector is calculated by using an adjacent image of the encoding target block, a disparity vector having high precision cannot be acquired in the case where a correlation between the adjacent image and the target block is low or a correlation with the adjacent image of a reference block referred as the disparity vector generated in Japanese Patent Application Laid-open No. 2005-159947 is low, and as a result, efficiency is not improved.